Fig. 2.

Hand detection. Upper: Hand detection in test videos using the final detector. (A) Green square shows detected hands. Warm colors indicate high detection score. (B and C) Examples of hands detected by appearance (B) and by context (C). Lower: Hand detection performance. (D) Detector trained from mover events only (magenta), and after one (blue) and three (green) iterations of cotraining by appearance and context. The top curve (red) shows performance obtained with full (manually) supervised learning. (E) Results after three iterations of training as before (green), without using context (blue) and without tracking (cyan). Error bars indicate SE across eight folds of cross-validation. Abscissa: Recall rate (fraction of detected hands out of all of the hands in the data); ordinate: precision (fraction of hands out of all detections).